Olefin hydrogenation using diimine pyridine complexes of Co and Rh

Article abstract of DOI:10.1016/j.molcata.2004.12.039

Square-planar cobalt diimine pyridine complexes LCoR (L = 2,6-[RN=CMe] ₂C₅H₃N; R = n-C₆H₁₃ for L^hex, 2,6-(i-Pr)₂C₆H₃ for L ^dip) are active in the hydrogenation of monosubstituted and disubstituted olefins; sterically more hindered trisubstituted olefins do not react. For the L^dipCo system, a diamagnetic hydride intermediate was observed, although a small amount of paramagnetic product is also formed upon reaction of L^dipCoR with H₂. DFT studies suggest a traditional hydrogenation cycle starting with LCoH, except that intermediate LCo(R)(H₂) transfers a hydrogen atom directly from H₂ to the alkyl group in a σ-bond metathesis step, without going through a discrete Co^III intermediate. Autoclave experiments show that conversion is not linear in catalyst intake. Diffusion limitation was ruled out as an explanation, and we propose a concentration-dependent catalyst decay. At low catalyst intake conversion rates up to 2 × 10⁴ (mol octene/mol Co/bar/h) can be reached. Reducing the steric bulk at the imine positions (L^dip → L^hex), or changing the metal from cobalt to rhodium, do not alter the activity or specificity of the hydrogenation much. For the L^hexCo and L^dipRh systems, no diamagnetic products corresponding to L^dipCoH were observed.

Full text of DOI:10.1016/j.molcata.2004.12.039

Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
Olefin hydrogenation using diimine pyridine complexes of Co and Rh
Quinten Knijnenburga, Andrew D. Hortonb, Harry van der Heijdenb, T. Martijn Kooistraa,
Dennis G.H. Hetterscheida, Jan M.M. Smitsa, Bas de Bruina,
Peter H.M. Budzelaara, ∗, Anton W. Gala
a
Metal-Organic Chemistry, University of Nijmegen, Toernooiveld 1, 6525 EDNijmegen, The Netherlands
b
CTCCT/1 Chemistry and Homogeneous Catalysis Shell International Chemicals B.V., P.O. Box 38000, 1030 BN Amsterdam, The Netherlands
Received 17 November 2004; received in revised form 16 December 2004; accepted 20 December 2004
Abstract
Square-planar cobalt diimine pyridine complexes LCoR (L = 2,6-[RN CMe]2C5H3N; R = n-C6H13 for Lhex, 2,6-(i-Pr)2C6H3 for Ldip) are
active in the hydrogenation of monosubstituted and disubstituted olefins; sterically more hindered trisubstituted olefins do not react. For the
LdipCo system, a diamagnetic hydride intermediate was observed, although a small amount of paramagnetic product is also formed upon
reaction of LdipCoR with H2. DFT studies suggest a traditional hydrogenation cycle starting with LCoH, except that intermediate LCo(R)(H2)
transfers a hydrogen atom directly from H2 to the alkyl group in a ␴-bond metathesis step, without going through a discrete CoIII intermediate.
Autoclave experiments show that conversion is not linear in catalyst intake. Diffusion limitation was ruled out as an explanation, and we
propose a concentration-dependent catalyst decay. At low catalyst intake conversion rates up to 2 × 104 (mol octene/mol Co/bar/h) can be
reached. Reducing the steric bulk at the imine positions (Ldip → Lhex), or changing the metal from cobalt to rhodium, do not alter the activity or
specificity of the hydrogenation much. For the LhexCo and LdipRh systems, no diamagnetic products corresponding to LdipCoH were observed.
© 2005 Elsevier B.V. All rights reserved.
Keywords: Hydrogenation; DFT calculations; EPR measurements; Cobalt; Diimine pyridine
1. Introduction
be made to perform analogously in other catalytic reactions
(e.g. hydrogenation). One possible approach would be the
use of a strong ␲-acceptor ligand that could stabilize low-
spin CoI while still allowing access to the CoIII oxidation
state.
Organometallic complexes of rhodium and (to a lesser ex-
tent) iridium play a prominent role in homogenous catalysis
[1,2]. They owe this role mainly to their ability to switch
between oxidation states +1 and +3, thus enabling a wide
range of reactions, in particular hydroformylation [3,4] and
hydrogenation [5,6]. The lighter group 9 congener cobalt has
a preference for the +2 and +3 oxidation states, which in gen-
eral prevents it from catalysing the same types of reactions.
However, cobalt carbonyls are active in hydroformylation,
suggesting that in combination with strong ␲-acceptor lig-
ands cobalt behaves more like rhodium. The low cost and
relatively low environmental impact would make cobalt a
very attractive alternative for rhodium and iridium, if it could
In earlier studies we have observed that the diimine pyri-
dine ligand has the ability to accept an electron from various
metal centers, forming ananionradical [7–10]. In recent work
on polymerisation with cobalt diimine pyridine complexes,
the groups of Gibson et al. [11] and ourselves [12] indepen-
dently discovered that en route to the active species, cobalt
passes through the +1 oxidation state. We have since observed
that with the same ligand the +3 oxidation state is still acces-
sible as well (vide infra). This prompted us to investigate
the possibility of using cobalt diimine pyridine complexes in
hydrogenation catalysis (part of the work has appeared as a
patent application [13]). Analogous Rh complexes [14] are
included here for comparison.
∗
Corresponding author. Tel.: +31 24 3652186; fax: +31 24 3553450.
E-mail address: budz@sci.kun.nl (P.H.M. Budzelaar).
1381-1169/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.molcata.2004.12.039

152
Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
2. Experimental
2.1. General
g-values and Co hyperfine couplings: g11 = 1.9935,
22 = 1.9985, g33 = 2.0185, A11 = 77.7 MHz, A22 = 31.2 MHz
and A33 = 26.8 MHz.
g
All manipulations were carried out under an atmosphere
of argon using standard Schlenk techniques or in a conven-
tional nitrogen-filled glovebox. Solvents were refluxed over
an appropriate drying agent and distilled prior to use. Hydro-
genation specificity experiments were carried out in NMR
tubesinC6D6 atroomtemperatureunderaninertatmosphere,
using ca. 5 mg of Co complex. NMR spectra were recorded
on Varian and Bruker spectrometers at ambient temperature.
X-band EPR spectra were recorded on a Bruker ER 220. The
spectra were simulated by iteration of the (an)isotropic g val-
ues, hyperfine coupling constants, and line widths.
2.3.3. LdipRhMe and H2
After addition of 2 ml of H2, all H NMR signals for
LRhMe disappeared. At T = 40 K in toluene, an EPR signal
was detected. Simulation yielded g = 1.9775, g = 2.0169,
1
ꢀ
⊥
A
ꢀ(Rh) = 40 MHz.
2.4. Hydrogenation of internal olefins
2.4.1. LdipCoCH2SiMe3 and internal olefin
Eight milligrams of LdipCoCH2SiMe3 and 8 mg 2-pentene
were dissolved in 0.5 g C6D6 and a 1H NMR spectrum was
recorded. Twominutesafter1 mlofH2 wasinjected, 1HNMR
showed a decrease in intensity of the vinylic protons around
5.4 ppm, indicating consumption of 2-pentene. Furthermore,
the signals for LdipCoCH2SiMe3 had disappeared, and, an-
other CoI complex had formed, which we identified as a CoI
n-pentyl species on basis of the ligand shifts (which resem-
bled earlier reported CoI alkyl species [12]) and the multiplet
for the n-pentyl ␤-hydrogens at −0.8 ppm. After addition of
an excess of H2, 1H NMR showed no traces of vinylic signals,
indicating that all 2-pentene had been converted to pentane.
Co was shown to be present in the form of LdipCoH.
The ligands Ldip [15] and Lhex [10] and the
complexes LdipCoCl2 [15], LhexCoCl2 [10], LdipRhCl
[16], LdipCoCH2SiMe3 [12], LhexCoCH2SiMe3 [10] and
L
dipRhMe [10] were synthesised following literature proce-
dures. All other starting materials were obtained commer-
cially and used as received.
2.2. LdipCo(CN)3
To 307 mg LdipCoCl2 (0.50 mmol) and 101 mg NaCN
(2.06 mmol; 4.12 equiv.) 25 ml chloroform was added. No
reaction had taken place even after 15 h of stirring. Addition
of 2 ml H2O resulted in a colour change from brown to red.
Addition of a layer of hexane resulted in the precipitation of
2.4.2. Selectivity of hydrogenation by LdipCoCH2SiMe3
for ␣-olefin in the presence of trisubstituted olefin
a dark red solid, which according to H NMR consisted of
Eight milligrams of LdipCoCH2SiMe3 was dissolved in
0.5 g of C6D6, and 5 ml H2 was added, resulting in the forma-
tion of LdipCoH as shown by 1H NMR. Subsequently eight
equivalents of 7-methyl-1,6-octadiene (MOD) were added.
1H NMR showed that the signals for LdipCoCH2SiMe3
had disappeared and a Co-alkyl complex had been formed
through insertion of the terminal double bond of MOD into
the Co H bond (the ligand signals were again comparable
to earlier observed alkyl species [12], and a multiplet for the
alkyl ␤-hydrogens was observed at −0.8 ppm). After addition
of more H2, all terminal double bonds were hydrogenated,
whereas the internal double bonds were still intact (only the
signal at 5.3 ppm remained in the vinylic region, whereas the
ones at 5.0 and 5.7 ppm associated with the terminal double
bond had disappeared), and again LdipCoH was left.
1
a mixture of 65% LdipCo(CN)3 and 35% free Ldip. A sim-
ilar experiment produced some crystals suitable for X-ray
structure determination (see below). Attempts to obtain ana-
lytically pure samples by crystallization failed, so we do not
have satisfactory elemental analysis data for this complex.
LdipCo(CN)3: H NMR (200 MHz, CDCl3): δ 8.61 (t,
1
3JHH 8 Hz, 1H, Py-H4), 8.34 (d, JHH 8 Hz, 2H, Py-H3,5),
3
3.36 (sept., 3JHH 6.5 Hz, 4H, iPr-H), 2.46 (s, 6H, N C-Me),
3
i
3
1.47 (d, JHH 6.5 Hz, 12H, Pr-Mea), 1.01 (d, JHH 6.5 Hz,
12H, iPr-Meb). IR (KBr): 2139 cm−1 (sh).
2.3. Reaction of metal-alkyl species with H2
2.3.1. LdipCoCH2SiMe3 and H2
This reaction was reported earlier to produce LdipCoH
[10]. The samples also show a weak EPR signal. Sim-
ulation yielded g = 2.0285, g = 1.996, A⊥(N) = 10 MHz,
2.4.3. LhexCoCH2SiMe3 and internal olefin
LhexCoCH2SiMe3 (6.9 mg) and 8 ␮l 2-hexene were dis-
solved in 2 ml hexane. To this solution 5 ml H2 was added.
1H NMR showed no traces of vinylic protons.
⊥
ꢀ
A
ꢀ(Co) = 30 MHz (see Fig. S1). At room temperature, a giso
value of 2.0112 was observed; we could not distinguish be-
tween a nitrogen superhyperfine coupling of 58.4 MHz or a
cobalt hyperfine coupling of 22.4 MHz.
2.4.4. Selectivity of hydrogenation by LhexCoCH2SiMe3
for ␣-olefin in the presence of trisubstituted olefin
2.3.2. LhexCoCH2SiMe3 and H2
Four milligrams of LhexCoCH2SiMe3 and 10 ␮l 7-methyl-
1,6-octadiene (MOD) were dissolved in 2 ml hexane. Ten
millilitres of H2 was added. 1H NMR showed traces of
the trisubstituted alkene bond, and other internal alkene
1
After addition of 2 ml of H2, all H NMR signals for
LhexCoCH2SiMe3 disappeared. At T = 40 K in toluene, an
EPR signal was detected. Simulation yielded the following

Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
153
Table 1
bonds (due to isomerisation products) but no monosubsti-
tuted alkene bond.
Crystallographic data for LdipCo(CN)3
Compound
LdipCo(CN)3
2.4.5. LdipRhMe and internal olefin
Crystal colour
Crystal shape
Crystal size (mm3)
Empirical formula
Formula weight
Temperature (K)
Transparent dark red
Rough platelet
Five milligrams of LdipRhMe was dissolved in 0.6 ml
C6D6 and placed in an NMR tube equipped with a sep-
tum. To the solution, 5 ␮l of a mixture of trans-2-hexene
0.38 × 0.11 × 0.03
C
36H43CoN6·4H2O
690.76
293(2)
1
and cis-2-hexene was added after which a H NMR spec-
˚
trum was recorded. One millilitre of hydrogen was injected,
the solution was stirred and a second 1H NMR spectrum was
recorded, which revealed that no vinylic protons were left,
indicating that all hexene had been converted to hexane.
Radiation/wavelength (A)
Crystal system, space group
Unit cell dimensions
MoK␣ (graphite mon.)/0.71073
Triclinic, P-1
˚
◦
a = 15.409(3) A, α = 108.592(15)◦
˚
b = 15.472(4) A, β = 101.604(15)◦
˚
c = 17.244(3) A, γ = 105.549(15)
3
˚
Volume (A )
3563.9(13)
2.4.6. Selectivity of hydrogenation by LdipRhMe for
␣-olefin in the presence of trisubstituted olefin
105275 reflections
Z, calculated density (mg/m3)
Absorption coefficient (mm−1
Diffractometer/scan
4, 1.287
0.528
Selectivity of hydrogenation by LdipRhMe for ␣-olefin
in the presence of trisubstituted olefin was determined fol-
lowing a similar procedure, which revealed that the terminal
alkene bond was hydrogenated (vinylic protons at 5.0 and
5.7 ppm had disappeared), whilst the trisubstituted one was
not (5.3 ppm remained).
)
Nonius KappaCCD with area detector
θ and ω scan
1472
F(0 0 0)
θ range for data collection (◦)
Index ranges
4.99–25.00
−18 ≤ h ≤ 18,
−18 ≤ k ≤ 18,
−20 ≤ l ≤ 20
Reflections collected/unique
Reflections observed
Absorption correction
Refinement method
Computing
105275/12462 [R(int) = 0.0951]
8113 [Io > 2σ(Io)]
2.5. High pressure hydrogenation procedures
SADABS multiscan correction [31]
Full-matrix least-squares on F2
SHELXL-97 [32]
12462/0/835
1.064
All products were identified and quantified by means of
GC–MS.
Data/restraints/parameters
Goodness-of-fit on F2
SHELXL-97 weight parameters
Final R indices [I > 2σ(I)]
R indices (all data)
0.0710, 3.4334
2.5.1. Mechanically stirred reactor
R1 = 0.0571, wR2 = 0.1362
R1 = 0.1063, wR2 = 0.−13600
The precatalyst was weighed into a Schlenk flask, which
was then flushed with argon. TIBA (1 M in hexane) and 40 ml
1-octene were added and this reaction mixture was brought
into a 100 ml pre-flushed steel autoclave. The mixture heated
and maintained at 50 ◦C by means of a thermocouple. Twenty
bar of hydrogen pressure was adjusted for 10 min. The reac-
tion mixture was stirred mechanically by means of a gas-
impelling stirrer.
˚
Largest diff. peak and hole
1.079 and −0.286 e A
at the unrestricted b3-lyp level [22–25] employing the stan-
dard SV(P) basis sets [26] and a small-core pseudopotential
for Co [27]. Numerical or analytical second derivatives were
calculated for all stationary points, and thermal corrections
(enthalpy and entropy, 323 K, 20 bar) were calculated from
these to produce the final relative free energies mentioned in
the text.
2.5.2. Magnetically stirred reactor
¨
A 100 ml glass autoclave (miniclave, Buchi AG) was
baked out at 150 ◦C and cooled to ambient temperature in
a nitrogen atmosphere. TIBA (0.5 ml 1 M in hexane), the cat-
alyst and 1-octene (10 ml, solvent as well as substrate) were
introduced in that order. The reactor was heated to 50 ◦C in an
oil bath, after which a constant 5 bar hydrogen pressure was
adjusted for 10 min. The reaction mixture was stirred with
a magnetic stirring bar at 200 rpm. To measure the decay of
activity, experiments were carried out under a 2-bar hydro-
gen pressure in the same autoclave setup, enabling sampling
through a septum without stopping the experiment.
2.7. X-ray structure determination for LdipCo(CN)3
The crystal data and a summary of the data collection and
structure refinement are given in Table 1 (for further details,
see Tables S2–S5 of the supplementary data). The structure
was solved by the PATTY option [28] of the DIRDIF pro-
gram system [29]. All nonhydrogen atoms were refined with
anisotropic temperature factors. The hydrogen atoms were
placed at calculated positions, and refined isotropically in
riding mode. The structure is shown in Fig. 1 [30]. The struc-
ture consists of two independent molecules. In addition, four
significant, isolated electron densities were found in the dif-
ference Fourier map. These were treated as disordered, par-
tially occupied water molecules. No hydrogen atoms could
be located for these oxygen atoms, so no hydrogen bonds
could be detected which would support this assignment. This
2.6. Calculations
Density functional calculations were performed with the
TURBOMOLE program [17–19] in combination with the
OPTIMIZE routine of Baker and co-workers [20,21]. All
relevant minima and transition states were fully optimised

154
Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
3.2. Synthesis of LdipCo(CN)3
This product was not used in hydrogenation experiments,
but is included here to demonstrate that the +3 oxidation
state is accessible for Co in its complexes with Ldip. The
reaction of LdipCoCl2 with NaCN in water/chloroform
under air produced LdipCo(CN)3. Layering with hexane
produced crystals that were suitable for X-ray diffrac-
tion. Fig. 1 shows one of the two independent molecules
present in the unit cell. The CoIII center atom has a
distorted octahedral geometry with an N1 Co N3 angle
of 162.66(12)◦. This distortion is imposed by the ligand
geometry and is also observed for other octahedral cobalt
complexes bearing this type of ligand [7]. Furthermore, the
two apical cyanides are bent away from the equatorial one,
presumably due to steric repulsion with the bulky diisopropy-
lphenyl groups, resulting in angles C5 Co C6 = 172.7(2)◦,
Co C5 N5 = 170.6(4)◦, Co C6 N6 = 167.0(4)◦ (com-
pare with Co C4 N4 = 175.2(4)◦). The Co C distances
(Co C4 = 1.894(4), Co C5 = 1.906(5), Co C6 = 1.898(5))
are similar to those found in other low-spin Co(III) cyanides
[33–36]. The IR spectrum shows a single, sharp IR band
at 2139 cm−1, which presumably represents coincident
␯as(CN)ax and ␯(CN)eq transitions (the ␯s(CN)ax transition
is expected to have a very low intensity). In related amine
complexes of Co(CN)3, ␯(CN) tends to be slightly smaller
(2127 cm−1 for a fac complex [35]; 2129/2133/2145 cm−1
for a mer complex [34]).
Fig. 1. ORTEP image of the X-ray structure of LdipCo(CN)3, drawn at 30%
probability; hydrogen atoms are omitted for clarity.
means that the chemical nature of these atoms is questionable.
3
˚
Finally, avoidof60 A wasfoundinthestructureintheneigh-
bourhood of O103. The SQUEEZE procedure showed no net
electron density in this void [30] although there are some
low residual electron densities in the difference Fourier map.
Geometrical calculations [30] revealed neither unusual ge-
ometric features, nor unusual short intermolecular contacts.
The calculations revealed no higher symmetry and no further
solvent accessible areas.
3. Results and discussion
3.1. Synthesis and characterisation of Co(I) complexes
3.3. Hydrogenation with LdipCo complexes [37]
The ligands Ldip and Lhex (Fig. 2) are prepared from the
condensation of two equivalents of the appropriate aniline or
amine with diacetylpyridine. For cobalt, the dichloride com-
plexes are synthesised in THF under air. The reduction and
subsequent alkylation from LhexCoCl2 to LhexCoCH2SiMe3
is achieved in one step using two equivalents of LiCH2Me3
[10]. LdipCoCH2SiMe3 [12] and LdipRhMe [14] were
synthesised according to literature procedures.
Reaction of LdipCoCH2SiMe3 with H2 produces a dia-
magnetic complex, which we believe to be the hydride
LdipCoH on the basis of 1H NMR [10].
Monosubstituted olefins easily insert into the Co H bond
of this hydride to form the corresponding alkyl derivatives.
Upon addition of more H2, the alkyl group is hydrogenated
off as an alkane and the hydride species is regenerated
(Scheme 1) [13]. Similarly, Gibson et al. [38] showed that in
the reaction of LdipCoMe with hydrogen a species is formed
believed to be LdipCoH, and which reacted with ethylene to
give LdipCoEt. 1,2-Disubstituted olefins can also insert and
be hydrogenated. For example, reaction of 2-pentene with
Fig. 2. Diiminepyridine ligands used.
Scheme 1. Hydrogenation cycle as followed by 1H NMR for LdipCoR.

Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
155
LdipCoH produces the n-pentyl complex, which then forms
pentane in the presence of excess hydrogen. This demon-
strates that alkyls can be isomerised at the cobalt center
prior to hydrogenation. Trisubstituted olefins, however, do
not react, showing that the hydrogenation activity is sensi-
tive to steric factors. This is illustrated by the hydrogenation
of 7-methyl-1,6-octadiene (MOD): only the monosubstituted
C C bond was hydrogenated. For catalytic hydrogenation, it
is more convenient to use LdipCoCl2 and activate it in situ
with TIBA (triisobutylaluminum), as discussed in more de-
tail below.
3.4. Hydrogenation with LhexCo complexes
In polymerisation with diiminepyridine complexes of Co
and Fe, steric shielding of the metal by bulky aryl sub-
stituents is necessary to obtain high molecular weight poly-
mers [15,39-41]. In order to test whether this type of shielding
is also needed for hydrogenation, complex LhexCoCH2SiMe3
[10], which carries n-hexyl groups on the imines, was
tested in hydrogenation. The product of the reaction be-
tween LhexCoCH2SiMe3 and H2 is not the expected hydride
LhexCoH but a paramagnetic species that shows a clear EPR
signal (see below) and is NMR-silent. Despite this difference
in behaviour between Ldip and Lhex complexes, catalytic hy-
drogenation using LhexCoCl2/TIBA shows an activity and
specificity similar to that of LdipCoCl2/TIBA.
Fig. 3. EPR of reaction product of LhexCoCH2SiMe3 and H2
(toluene glass, 40 K). Simulation: g11 = 1.9935, g22 = 1.9985, g33 = 2.0185,
A
11(Co) = 77.7 MHz, A22(Co) = 31.2 MHz and A33(Co) = 26.8 MHz.
at r.t. as seen for LdipCoR + H2 is indicative of mainly
ligand-centred radicals
It is remarkable that upon reaction with H2 LhexCoR
generates exclusively a paramagnetic complex, while for
LdipCoR the paramagnetic product is only a minor product.
For both li•gands, the paramagnetic complex is prob•a−bly a
low-spin L −Co ligand radical complex (in which L de-
notes the radical anion of L), where the unpaired electron
is delocalised to the metal to different extents. Acces to the
paramagnetic species seems to be promoted by the decreased
steric bulk of the Lhex ligand compared to Ldip. The exact
nature of these paramagnetic species cannot be derived from
the EPR spectra.
3.5. Hydrogenation with LdipRh complexes
LdipRhMe also reacts with H2 to form a paramagnetic
and NMR-silent product [14]. Catalytic hydrogenation using
LdipRhCl/TIBA shows a specificity with MOD similar to that
of the Co complexes.
The EPR spectrum of the paramagnetic product obtained
3.6. EPR measurements.
from LdipRhMe and H2 (toluene glass, 40 K) shows an ax-
ial signal indicative of an S = 1/2 species, of which g re-
The paramagnetic product of LhexCoCH2SiMe3 and H2
shows a clear EPR signal indicative of an S = 1/2 species.
The rhombic signal (Fig. 3) could be simulated satisfactorily
using only hyperfine coupling to Co.
ꢀ
veals superhyperfine coupling; simulation yields g = 1.9775,
ꢀ
g = 2.0169, Aꢀ(N) = 40 MHz [14]. Since the g-values of the
⊥
paramagnetic rhodium species are close to ge, it seems logical
to assume that the unpaired electron is mainly located on a lig-
and fragment, presumably in one of the pyridine 2,6-diimine
␲* orbitals. This would imply a formulation as L•−RhI or
L•−RhIIIH2.
On closer inspection we found that also in the reaction
between LdipCoCH2SiMe3 and H2, besides LdipCoH, some
paramagnetic product is formed (at most 5% according to a
1
quantitative H NMR experiment). In this case, interpreta-
tion of the EPR signal is less straightforward. The spectrum
in toluene glass (T = 40 K) shows a broad signal indicative
of axial symmetry (g = 2.0825, g = 1.997) (spectrum:
3.7. Hydrogenation activity
ꢀ
⊥
Fig. S1, in supplementary data). At room temperature,
the isotropic signal at g = 2.0112 (Fig. 4) reveals some
(super)hyperfine coupling; we cannot unambiguously
discriminate between a superhyperfine coupling to a single
nitrogen (Aiso(N) ≈ 58 MHz, Fig. 4A) and a cobalt hyperfine
coupling (Aiso(Co) ≈ 22 MHz, Fig. 4B), although the former
gives a slightly better fit. The observation of isotropic signals
Rates of hydrogenation were determined with 1-
octene as substrate and solvent. In general, the dichloride
(cobalt) or monochloride (rhodium) precursors were used
(0.2–25 ␮mol) in combination with a large excess of TIBA
(typically 0.5 mmol) as activator and scavenger. TIBA (1 M
in hexane) was first added to the solid precursor, since the
activated catalyst dissolves in hexane (and 1-octene). For low

156
Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
Fig. 4. EPR of reaction product of LdipCoCH2SiMe3 and H2 (room temperature). Simulation: giso = 2.0112; (A) Aiso(N) = 58.4 MHz; (B) Aiso(Co) = 22.4 MHz.
intake experiments this solution was further diluted and extra
TIBA was added in order to maintain the same scavenger
concentration. Control experiments using pre-formed metal
alkyls showed that in situ activation by TIBA was effective.
Hydrogenation rates at low intake (ca. 0.5 ␮mol catalyst,
glass autoclave, magnetic stirring, 5 bar H2) showed con-
version of 1-octene by LdipCoCl2, LhexCoCl2 and LdipRhCl
with TIBA in the order of 10,000 mol/mol catalyst/bar/h. For
LdipCoCl2, conversion was found to vary linearly with hydro-
gen pressure in the range of 5–50 bar (see Table S1 and Figs.
S2–S4 in supplementary data for details of all experiments).
However, conversion did not depend linearly on catalyst
intake. The possibility that this might be due to diffusion
limitation was ruled out by experiments using LdipCoCl2 in a
mechanically stirred reactor at 20 bar. In this setup, hydrogen
was added efficiently directly into the octene by means of a
Fig. 5. Conversion versus intake (LdipCoCl2, mechanical stirring, 20 bar
H2).
gas-impelling stirrer. Stirring at full or half capacity gave the
same conversion.
Fig. 5 shows the results for four separate experiments us-
ingvaryingamountsofcatalyst(furtherresultsunderavariety
of stirring conditions and pressures are listed in Table S1 of
the supplementary data). The graph clearly shows that higher
catalyst intake leads to lower turnovers. Such behaviour is
sometimes observed if the catalyst is sensitive to the product.
In the present case, where the product is a saturated alkane,
this is highly unlikely. Therefore, we suspect that a bimolec-
ular catalyst deactivation may be responsible [42].
For LdipCoCl2, the decay of activity with time was mea-
sured as well. The magnetically stirred setup was loaded with
a large amount of catalyst, and a hydrogen pressure of two bar
was maintained for one hour. This low pressure enabled sam-
pling through a septum. Conversion was monitored by taking
Fig. 6. Conversion vs. time for hydrogenation of 1-octene with LdipCoCl2,
fitted assuming second-order (solid line) or first-order (dashed line) catalyst
decay.
GC samples at 10-min intervals (Fig. 6). Some form of cata-
lyst decay is obvious from these data. The solid and dashed
lines in the figure are least-squares fits assuming second-

Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
157
order and first-order catalyst decay, respectively. The fit is
marginally better for the second-order model, but the data
are not good enough to allow discrimination between the two
models.
as a ␴-bond metathesis reaction rather than an oxidative
addition/reductive elimination sequence. Thus, even though
we have demonstrated that for Co coordinated to Ldip the
+3 oxidation state is accessible, it is apparently not needed
in this hydrogenation cycle. The rate-determining step of
the cycle is the ␴-bond metathesis step, with an activation
energy (relative to LCoEt) of about 21 kcal/mol [44].
The Co(0) cycle starts with LCo. Of the two possible ini-
tial reactions, complexation of ethylene or H2, the former is
strongly preferred. After ethylene coordination, binding of
H2 to form LCo(C2H4)(H)2 is strongly endothermic. Once
this complex has been formed, both ethylene insertion and
subsequent reductive ethane elimination have low barriers.
The rate-determining step of this cycle is ethylene insertion,
with a barrier (relative to LCo(C2H4) 13) of ca. 30 kcal/mol,
i.e. somewhat higher than found for the CoI cycle.
3.8. Theoretical studies
In order to shed more light on the possible mechanism(s)
of the hydrogenation catalysis, we have studied the reaction
using DFT methods. Both CoI and Co(0) cycles were consid-
ered. Fig. 7 shows the free energy profiles calculated for the
two cycles using a simple ligand model (H substituents only)
and ethylene as the model substrate.
The relative free energies for the species occurring in the
hydrogenation cycles of Fig. 7 are listed in Table 2 (total
energies and low frequencies are listed in Table S6).
The CoI cycle is straightforward. Ethylene coordinates
weakly to LCoH (1) and then inserts with a low barrier
to give CoI alkyl 4. This then coordinates H2, but does
not cleave the H H bond [43]. Ethane formation proceeds
through direct transfer of one of the H2 hydrogens of
5 to the ethyl group, i.e. this step should be considered
Both cycles require approximately the same amount of
space around the metal, so steric effects should be simi-
lar. Based on the computational results, it seems that the
CoI cycle is likely to be the most important one, which at
least for Ldip is in agreement with the experimental results
[45].
Fig. 7. Free energy profiles (kcal/mol) for CoI/CoIII and Co(0)/CoII hydrogenation cycles.

158
Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
Table 2
Acknowledgements
Relative free energies for species in the two hydrogenation cycles
Complex
Erel (kcal/mol)
¨
We thank Prof. D. Vogt and Dr. C. Muller for helpful dis-
CoI cycle
cussion on the subject of diffusion limitation, and for the use
of the mechanically stirred autoclave setup for high-pressure
hydrogenation experiments. We are grateful to Dr. M.G. Ni-
jkamp and Drs. M.M.P. Grutters for their assistance with the
use this equipment. We thank Dr. F. Neese for a copy of his
EPR simulation program.
1
2
3
4
5
6
7
8
LCoH
(0)a
LCoH(C2H4)
LCo(H···C2H4)
LCoEt
LCo(Et)(H2)
LCo(Et)(H)2
LCo(Et···H···H)
LCoH(C2H6)
LCoH + C2H6
−2.76
0.30
−15.41a
−0.58b
c
–
5.87
−19.12
−26.28
9
Co(0) cycle
Appendix A. Supplementary data
10
11
12
13
14
15
16
17
18
19
20
a
LCo
LCo(H2)
LCo(H)2
LCo(C2H4)
(0)
4.10
Supplementary data associated with this article can
be found, in the online version, at 10.1016/j.molcata.
2004.12.039.
d
–
−10.84
e
LCo(C2H4)(H)2
LCo(H···H···C2H4)
LCo(H)(Et) (H apical)
LCo(H)(Et) (Et apical)
LCo(H···Et)
–
EPR spectrum measured at 40 K of the reaction between
18.76
−2.20
0.60
L
dipCoCH2SiMe3 and H2 (Fig. S1), full experimental details
about additional hydrogenation runs (Table S1, Figs. S2–S4),
crystallographic data for LdipCo(CN)3 (Tables S2–S5), to-
tal energies, thermal corrections and lowest frequencies for
species studied theoretically (Table S6)
Crystallographic data (excluding structure factors) for
LdipCo(CN)3 have been deposited with the Cambridge
Crystallographic Data Centre as supplementary publication
CCDC 237217. Copies of available material can be obtained,
free of charge, on application to the Director, CCDC, 12
Union Road, Cambridge CB2 1EZ, UK, (fax: +44 1223
336033 or e-mail: teched@chemcrys.cam.ac.uk).
1.33
LCo(C2H6)
LCo + C2H6
−14.81
−26.28
Singlet biradical, prefers a broken-symmetry solution. For details see
Ref. [10].
b
Regular local minimum on the restricted b3-lyp surface. However, only
about 0.2 kcal/mol above this minimum in the direction of H2 dissociation,
the broken-symmetry solution becomes more stable, and then the energy
decreases monotonically to 4.
c
Not a local minimum, collapses to H2 complex 5.
Very shallow minimum, ca. 7.7 kcal/mol above 11, with a barrier
d
<0.1 kcal/mol for collapse back to 11.
e
Not a local minimum, but on a very flat plateau ca. 1.3 kcal/mol below
15.
References
[1] B. Cornils, W.A. Herrmann, Applied Homogenous Catalysis with
Organometallic Compounds, Weinheim, 1996.
[2] G.W. Parshall, S.D. Ittel, Homogenous Catalysis, second ed., John
Wiley, New York, 1992.
4. Conclusions
The square planar CoI alkyl species that were recently
discovered as intermediates in the activation pathway for
the polymerisation of ethylene are also active catalysts for
the hydrogenation of olefins. Introduction of less bulky sub-
stituents at the imine nitrogens does not affect this catalyst
activity much, in contrast to the ligand effects observed in
polymerisation [15,39–41]. Establishing the active species
is difficult for any catalytic reaction. In the case of the
LdipCo catalyst, stoichiometric experiments have demon-
strated both olefin insertion into the hydride and hydrogenol-
ysis of the metal-alkyl bond, supporting a cycle involving
LdipCoH. However, the fact that also a small amount of
paramagnetic product (<5%) is formed upon reaction with
H2 means that the possibility of a paramagnetic species be-
ing active as well cannot be ruled out. Calculations are not
conclusive but indicate a cycle involving LdipCoH is rea-
sonable. This shows that, given the right ligand environ-
ment, Co can indeed be made to display “rhodium-like”
catalysis.
[3] Ref. [1], section 2.1.1.
[4] Ref. [2], section 5.4.
[5] Ref. [1], section 2.2.
[6] Ref. [2], chapter 3.
[7] B. De Bruin, E. Bill, E. Bothe, T. Weyhermuller, K. Wieghardt,
Inorg. Chem. 39 (2000) 2936.
[8] P.H.M. Budzelaar, B. De Bruin, A.W. Gal, K. Wieghardt, J.H. Van
Lenthe, Inorg. Chem. 40 (2001) 4649.
[9] H. Sugiyama, G. Aharonian, S. Gambarotta, G.P.A. Yap, P.H.M.
Budzelaar, J. Am. Chem. Soc. 124 (2002) 12268.
[10] Q. Knijnenburg, D. Hetterscheid, T.M. Kooistra, P.H.M. Budzelaar,
Eur. J. Inorg. Chem. 6 (2004) 1204.
[11] V.C. Gibson, M.J. Humphries, K.P. Tellmann, D.F. Wass, A.J.P.
White, D.J. Williams, Chem. Commun. 21 (2001) 2252.
[12] T.M. Kooistra, Q. Knijnenburg, J.M.M. Smits, A.D. Horton, P.H.M.
Budzelaar, A.W. Gal, Angew. Chem. Int. Ed. Engl. 40 (2001)
4719.
[13] A.D. Horton, Q. Knijnenburg, H. van der Heijden, P.H.M. Budzelaar,
A.W. Gal, WO 03042131 (2003); EP 2001-204376 (2001).
[14] T.M. Kooistra, D.G.H. Hetterscheid, E. Schwartz, Q. Knijnenburg,
P.H.M. Budzelaar, A.W. Gal, Inorg. Chim. Acta 357 (2004) 2945.
[15] G.J.P. Britovsek, M. Bruce, V.C. Gibson, B.S. Kimberley, P.J. Mad-
dox, S. Mastroianni, S.J. McTavish, C. Redshaw, G.A. Solan, S.

Q. Knijnenburg et al. / Journal of Molecular Catalysis A: Chemical 232 (2005) 151–159
159
Stro¨mberg, A.J.P. White, D.J. Williams, J. Am. Chem. Soc. 121
(1999) 8728.
[31] G.M. Sheldrick, SADABS. Program for Empirical Absorption Cor-
rection, University of Gottingen, Germany, 1996.
[16] H.F. Haarman, J.M. Ernsting, M. Krananburg, H. Kooijman, N. Veld-
man, A.L. Spek, P.W.N.M. van Leeuwen, K. Vrieze, Organometallics
16 (1997) 887.
[17] R. Ahlrichs, M. Ba¨r, M. Ha¨ser, H. Horn, C. Ko¨lmel, Chem. Phys.
Lett. 162 (1989) 165.
[32] G.M. Sheldrick, SHELXL-97. Program for the refinement of crystal
structures, University of Gottingen, Germany, 1997.
[33] M. Ishii, M. Umehara, M. Nakahara, Bull. Chem. Soc. Jpn. 60 (1987)
125.
[34] H. Chun, E.M. Maes, R.S. Czernuszewicz, I. Bernal, Polyhedron 20
(2001) 2597.
[18] O. Treutler, R. Ahlrichs, J. Chem. Phys. 102 (1995) 346.
[19] R. Ahlrichs, M. Ba¨r, H.-P. Baron, R. Bauernschmitt, S. Bo¨cker, M.
Ehrig, K. Eichkorn, S. Elliott, F. Furche, F. Haase, M. Ha¨ser, C.
Ha¨ttig, H. Horn, C. Huber, U. Huniar, M. Kattannek, A. Ko¨hn, C.
[35] J.Y. Yang, M.P. Shores, J.J. Sokol, J.R. Long, Inorg. Chem. 42 (2003)
1403.
[36] H. Chun, I. Bernal, Cryst. Growth Des. 1 (2001) 143.
[37] After the present work was submitted, an experimental/theoretical
study of the CoI path of olefin hydrogenation with LdipCo ap-
peared K.P. Tellmann, M.J. Humphries, H.S. Rzepa, V.C. Gibson,
Organometallics 23 (2004) 5503.
[38] V.C. Gibson, K.P. Tellmann, M.J. Humphries, D.F. Wass, Chem.
Commun. 20 (2002) 2316.
[39] B.L. Small, M. Brookhart, A.A. Bennett, J. Am. Chem. Soc. 120
(1998) 4049.
¨
Ko¨lmel, M. Kollwitz, K. May, C. Ochsenfeld, H. Ohm, A. Scha¨fer,
U. Schneider, O. Treutler, K. Tsereteli, B. Unterreiner, M. von
Arnim, F. Weigend, P. Weis, H. Weiss Turbomole Version 5, January
2002. Theoretical Chemistry Group, University of Karlsruhe.
[20] PQS version 2.4, 2001. Parallel Quantum Solutions, Fayetteville,
Arkansas, USA (the Baker optimizer is available separately from
PQS upon request).
[21] J. Baker, J. Comput. Chem. 7 (1986) 385.
[22] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785.
[23] A.D. Becke, J. Chem. Phys. 98 (1993) 1372.
[24] A.D. Becke, J. Chem. Phys. 98 (1993) 5648.
[40] G.J.P. Britovsek, V.C. Gibson, B.S. Kimberley, P.J. Maddox, S.J.
McTavish, G.A. Solan, A.J.P. White, D.J. Williams, Chem. Commun.
18 (1998) 849.
[25] All calculations were performed using the Turbomole functional “b3-
lyp”, which is not identical to the Gaussian “B3LYP” functional.
[26] A. Scha¨fer, H. Horn, R. Ahlrichs, J. Chem. Phys. 97 (1992)
2571.
[27] D. Andrae, U. Haeussermann, M. Dolg, H. Stoll, H. Preuss, Theor.
Chim. Acta 77 (1990) 123.
[41] B.L. Small, M. Brookhart, J. Am. Chem. Soc. 120 (1998) 7143.
[42] Catalyst preparation times were kept constant. If catalyst decompo-
sition was first-order, initial rates should not be affected by it.
[43] The ethyl-dihydride complex LCo(Et)(H)2 was found not te be a lo-
cal minimum; it rearranges spontaneously to the dihydrogen complex
LCo(Et)(H2).
[28] P.T. Beurskens, G. Beurskens, M. Strumpel, C.E. Nordman, in: J.P.
Glusker, B.K. Patterson, M. Rossi (Eds.), Patterson and Pattersons,
Clarendon Press, Oxford, 1987, p. 356.
[29] P.T. Beurskens, G. Beurskens, W.P. Bosman, R. de Gelder, S. Garcia-
Granda, R.O. Gould, R. Israel, J.M.M. Smits, DIRDIF-96. A com-
puter program system for crystal structure determination by Patterson
methods and direct methods applied to difference structure factors,
Crystallography Laboratory, University of Nijmegen, The Nether-
lands, 1996.
[44] This should be considered as a rough estimate only. LCoEt prefers
a broken-symmetry ub3-lyp solution. In the present work we have
used the ub3-lyp energies throughout, but for an accurate energy
estimate one would need a “true singlet” energy of this and similar
species. See, e.g. Ref. [8].
[45] One could assert that the formation of paramagnetic species from
LhexCoR and LdipRhR and H2 argues against an MI cycle for
these two precursors. However, if paramagnetic (M(0) or MII) species
are easily formed from MI precursors, the reverse reaction is also
possible, so diamagnetic cycles are still possible even for these two
systems.
[30] A.L. Spek, PLATON. A Multipurpose Crystallographic Tool, Utrecht
University, Utrecht, The Netherlands, 2003.